Stress singularity is usually observed in practical stress analysis, and it may lead to the deterioration of the convergence rate of the preconditioned conjugate gradient method (PCG). Until now, parallel PCG has been used without considering the practical issues; therefore, the deterioration was induced in the problem in which stress singularity was observed. The convergence acceleration method was developed in parallel PCG based on controlled domain decomposition. The acceleration method can be achieved by considering the spatial locality of stress singularity and preconditioning of conjugate gradient method. The convergence acceleration method reduces the 15% of iteration of PCG as a maximum, where the singular area is included in one domain.

View full abstract

Innovative thin film technology to realize the finer electric devices needs to understand the atomic level process of film growth and its relationship to the film characterization. In this paper, the long film growth phenomena for a few micro-second order with the short severe collisions by incident particles are analyzed by the proposed hybridized atomistic modeling. This method combined molecular dynamics (MD) with kinetic Monte Carlo (KMC) can directly treat two types of events of deposition and diffusion, which have quite different time scales. The solutions suggest that the large incident kinetic energy of deposited atoms compatible to the realistic physical vapor deposition (PVD) impels to fluctuate the equilibrium on Al (111) surface very drastically and affects the atomic level surface morphology. It is found that the faster incident atoms with 1.0×10⁴ m/s can make the smoother surface than those with the velocity of 1.0×10³ m/s. This is due to much activated atomic migration, which can be realized only by MD.

View full abstract

This paper describes both the development of a new low-cost optimization method for the optimal design of laminated composite structures, and a result of applying this method to the minimum weight design of a hat-stiffened composite panel subject to a buckling constraint. The new method is based on two levels of optimization. In the low level optimization, the fractal branch and bound method is used for a stacking sequence optimization. In the high level optimization, the particle swarm optimization technique is used for optimizing configurations of the hat-stiffened panel. Moreover, a Kriging-based response surface is used to find the global optimum, and not just to approximate an objective function value. For this purpose, we use the ‘expected improvement’ (EI) criterion defined in the Kriging model. The validity of the proposed method is examined and the results show that the method provides valuable solutions with good precision and a low computational cost.

View full abstract